Does FDI promote regional development? Evidence …in 2008 manufacturing accounted for some 33% of the total stock of inward FDI, 3 Ambiguity also exists with regard to the location

EASTERN JOURNAL OF EUROPEAN STUDIES Volume 1, Issue 2, December 2010 139

Does FDI promote regional development?

Evidence from local and regional productivity

spillovers in Greece

Vassilis MONASTIRIOTIS* and Jacob A. JORDAAN

Abstract

Studies on the productivity spillovers of FDI have concentrated on the national-

sectoral level. As a result, little is known about the impact of FDI on absolute

and relative regional economic performance. In this paper we examine this issue

by relying on a unique dataset of over 20,000 Greek firms for the period 2002-

2006 covering all sectors of economic activity. We examine the spatial

distribution of foreign-owned firms in the country and analyse the effect that

their presence – at the local, regional and national levels – has on the

productivity of domestic firms. We find strong evidence suggesting that foreign-

owned firms self-select into regions and sectors of high productivity. Net of this

selection effect, the impact of foreign presence on domestic productivity is

negative – although at the very local level some positive spillover effects are

identifiable. The bulk of the effects concentrate in non-manufacturing activities,

high-tech sectors, and medium-sized high-productivity firms. Importantly, this

effect is not constant across space however. Productivity spillovers tend to be

negative in the regions hosting the main urban areas in the country but positive

in smaller and more peripheral regions. In this way, despite the tendency of FDI

to concentrate in a limited number of areas within the country – those of the

highest level of development – the externalities that FDI activity generates to the

local economies appear to be of a rather equilibrating character.

Key words: regional development, FDI, productivity spillovers, Greece, spatial

heterogeneity

JEL classification: F23, R11, C23, O12

* Vassilis Monastiriotis is senior lecturer at European Institute, London School of Economics,

Houghton Street, WC2A 2AE London, UK; e-mail: [email protected]. Jacob A. Jordaan is assistant professor at Department of Economics, VU University Amsterdam,

1105 de Boelelaan, 1081 HV, Amsterdam, the Netherlands; e-mail: [email protected].

140 Vassilis MONASTIRIOTIS and Jacob A. JORDAAN

1. Introduction

Foreign direct investment can be an important source of economic

development for recipient economies. FDI inflows strengthen capital

accumulation and job creation domestically, while they improve the fiscal and

external position of the recipient countries, thus helping finance government

expenditures that can further stimulate economic development (Caves, 2007;

Dunning and Lundan, 2008). Besides these macroeconomic effects, foreign

investments may have more direct effects on industrial activity and performance,

through their impact on the technology and productivity of domestic firms.

Owing to their technological and other advantages, foreign-owned firms can

generate significant externalities for the domestic economy. These can operate

mainly through three channels: pecuniary/demand spillovers, technological/

learning spillovers, and competition effects1.

There is now a large literature examining the magnitude and direction of

these effects. Reflecting its industrial and business economics origins, the

literature has typically focused at the national-sectoral level, as these spillovers

are assumed to operate along sectoral lines. As a consequence, only limited

attention has been paid to the spatial distribution of FDI spillovers and the

identification of region-specific effects accruing from FDI. Of the few cases that

have studied regional-level FDI spillovers, most have either treated the

geographical information as an additional dimension in their sample (Aitken and

Harrison, 1999; Blalock and Gertler, 2008) or have limited their focus to the role

that industrial clusters and agglomeration play for the realisation of FDI

spillovers (Driffield and Munday, 2001; Jordaan, 2005; De Propris and Driffield,

2006) – both without explicitly examining the geography of such spillovers.

Only very few studies have instead examined the process of diffusion of FDI

spillovers across space (Halpern and Muraközy, 2007; Haskel et al, 2007;

Jordaan, 2008) and, to our knowledge, only one has examined specifically the

issue of differentiation in the direction and magnitude of FDI spillovers across

space (Mullen and Williams, 2007). By implication therefore, there is a notable

gap in our knowledge and understanding of the overall impact of foreign

investment on the host country‟s regional economic structure and performance.

This paper makes an important contribution in this regard, by providing an

analysis of the location and productivity spillovers of FDI at the regional and

sub-regional (local) level and examining both the process of spatial diffusion of

1 See Blomstrom and Kokko (1998) and Aitken and Harrison (1999) for an early literature on these

issues and Smeets (2008) for an excellent review of these theoretical channels. It should be noted

that spillovers can also be negative (and empirically they appear to be so more often than not),

especially in cases where market-capturing by the foreign affiliates creates a „demand-siphoning‟

effect for the domestic firms which raises average production costs and lowers productivity.

Moran et al (2005) offer an extensive discussion of positive and negative productivity effects that

arise from FDI.

DOES FDI PROMOTE REGIONAL DEVELOPMENT? 141

FDI spillovers and the extent of differentiation of these spillovers across space.

We focus on the case of Greece, a country with a significantly skewed

geographical production structure (over-concentration of economic activity in a

few centres) and a low degree of FDI attraction. Owing to these features, Greece

appears as a particularly appealing empirical case – where the geographical

concentration of FDI is expected to be more intense and spatial differences in the

extent of spillovers more heightened. Our analysis sets to explore three main

questions. First, what has been the incidence and sectoral distribution of foreign

activity across the regional economies of Greece. Second, what is the effect of

foreign activity on the productivity of domestic firms in the country and how

localised is this effect. Third, whether and to what extent are FDI spillovers

regionally differentiated and conditioned upon specific firm-level, sectoral and

regional characteristics.

Our data refer to a firm-level panel of annual observations covering the

period 2002-2006, a period of relative stability and fast growth, starting after the

country‟s successful adoption of the EURO and ending before the first signs of

the global financial crisis. All data are derived from the Amadeus database

produced by Bureau van Dijk (BvD), which contains firm-level information on

turnover, fixed assets, employment, ownership and other variables of interest for

the majority of European countries, covering all sectors of economic activity.

For Greece, the Amadeus database contains some 27,000 firms per year, of

which just over 2,000 are foreign owned, representing an employment share of

7.78%.2 This dataset is unique in its detail and coverage and, to our knowledge,

it has not been used before for the case of Greece.

Our starting premise is that FDI spillovers are essentially and

fundamentally heterogeneous across space – not least because FDI itself tends to

be particularly concentrated, especially in locations of high productivity,

accessibility and industrial agglomeration. The scant existing empirical evidence

seems to support this claim, showing that industrial clustering (agglomeration) is

significant not only for attracting foreign firms (Guimaraes et al, 2000; Hilber

and Voicu, 2010) but also for determining the size of the realised productivity

spillovers (Driffield and Munday, 2001; Jordaan, 2005; De Propris and Driffield,

2006). There are also good theoretical reasons to expect FDI-induced

productivity spillovers to take a heterogeneous geographical manifestation –

although the effects can go either way. On the one hand, foreign investments

may be less able to generate positive spillovers in less developed regions

2 As is standard in the literature, we define a firm as foreign owned if at least 10% of its value is

owned by an ultimate owner who is resident or established outside the country. After excluding

thus defined foreign affiliates and observations with incomplete or erroneous information, our

estimating sample reduces to just over 20,000 firms (98,408 firms-specific observations in the

pooled sample) – bringing the average employment share of foreign-owned firms to just below

13%.


because their technological distance to local firms does not allow for potential

spillovers to be absorbed (the „absorptive capacity‟ argument – see Kokko,

1994; Damijan et al, 2001; Girma, 2005; Jordaan, 2009). On the other hand,

foreign investments may produce larger spillovers in less developed regions, as

firms there are typically less exposed to international competition and have thus

more to „learn‟ from the foreign-owned firms (the „scope for spillovers‟

argument – see Kinoshita, 2000; Merlevede and Schoors, 2005; Gersl et al,

2007; Monastiriotis and Alegria, 2011).3

In the remainder of this paper we set to explore the empirical validity of

these expectations by providing a holistic analysis of the incidence, geographical

scale and spatial differentiation of FDI-induced productivity spillovers across the

Greek regions. We examine the incidence and sectoral distribution of foreign

activity in section 2. Section 3 presents our econometric analysis, which explores

in detail the FDI productivity spillovers and the extent to which they diffuse

across space (how localised they are and how neighbouring FDI affects local

productivity). In section 4 we analyse the issue of spatial and functional

differentiation of these spillovers. We conclude with a short discussion of the

implications of our findings for both policy and the empirical literature on FDI

spillovers.

2. FDI in Greece and its regions

Historically, Greece has not been an important recipient of FDI. The

country embarked on a policy to encourage inward investments since the 1950s

and while FDI flows recorded an almost continuous growth (in absolute terms)

for decades (Louri et al, 2000), its total inward FDI stock was below 10% of

GDP in the 1990s and 2000s with annual FDI flows representing less than 10%

of total gross fixed capital formation in the country (UNCTAD, 2009). As a

result, Greece ranks persistently at the bottom of the international rankings of

FDI recipients and its FDI stock represents less than 1% of the total inward FDI

stock of the EU27. Moreover, it appears that the technology content of inward

FDI in Greece is also particularly low: according to data by the Bank of Greece,

in 2008 manufacturing accounted for some 33% of the total stock of inward FDI,

3 Ambiguity also exists with regard to the location of FDI across regions. Theory and empirical

evidence suggest that FDI tends to locate in national capitals and some highly accessible and

relatively developed regions, which “cream-off” the most productive foreign investors, thus

reinforcing existing spatial asymmetries in production structures and capabilities (Guimaraes et al,

2000; Cantwell and Iammarino, 2001; Resmini, 2008 – see Pantelidis and Nikolopoulos, 2008 for

the case of Greece). In theory, however, high-productivity foreign-owned firms may also choose to

locate in less developed regions, as part of a strategy to protect their technological advantages

from diffusing to their domestic competitors who are typically located in the more developed

regions of the host country.


almost two-thirds of which was in sectors producing consumer goods, with FDI

in the manufacture of capital goods representing a mere 0.8% of the total FDI

stock in the country.

This is consistent with findings of previous research, which has shown

that FDI is below the country‟s potential (Papazoglou, 2001; UNCTAD, 2004;

Kokkinou and Psycharis, 2004), predominantly of a market-seeking type

(Georganta et al, 1986; Georgopoulos and Preusse, 2006) and concentrating in

traditional sectors that are characterised by low technology and labour-intensive

production (Barrios et al, 2004). The low degree and quality of FDI in the

country has often been attributed to factors such as the extent of red-tape and

bureaucracy, high tax rates, poor infrastructure and a weak business and

macroeconomic environment (Apergis and Katrakylidis, 1998; Barbosa and

Louri, 2002; Filippaios and Kottaridi, 2004; Pantelidis and Nikolopoulos, 2008).

Previous research has also shown that FDI in Greece is highly

concentrated, along both sectoral and spatial lines (Dimelis et al, 2004; Bitzenis

et al, 2007). Indeed, together with manufacturing, three other sectors account

jointly for over 90% of the FDI stock in the country (financial services 30%,

transport and communications 15%, wholesale/retail trade 13%).4 Interestingly,

the Hotel and Restaurants sector, which includes the tourism industry, one of the

country‟s main comparative advantages, only accounts for 2% of total FDI

stock. FDI appears also particularly concentrated across space. Bank of Greece

data show that in the period 2000-2008 only 25 out of the 51 NUTS3 regions of

the country received any form of FDI, with 87% of FDI inflows going to the

prefecture of Attiki, where the national capital is situated, and the fifth highest

FDI recipient accounting for a mere 0.5% of total FDI inflows into the country

(€100m compared to a national figure of €18.8bn).

Thus, both along sectoral and geographical lines, the distribution of

foreign-owned activity in Greece is particularly skewed, with FDI being of an

important relative size in only a few sectors and regions. This is also revealed in

the data derived from the Amadeus database. Using this data, Figure 1 presents

the geography of FDI concentration (employment in foreign-owned firms as a

share of total employment in each region) at the NUTS3 level, averaged over the

2002-2006 period.

4 Data refer to the on the book value of investments derived from the Bank of Greece. In our data

we get a similar picture of concentration in these main sectors, although given that we use

employment than investment shares the ranking of the sectors is different, reflecting sectoral

differences in capital-labour ratios. The four main sectors account in our data for 85% of total

employment in foreign-owned firms (wholesale and retail trade: 37%; manufacturing: 26%;

transport and telecommunications: 14%; financial and business services: 8%).


Figure 1. Presence of foreign-owned activity in the Greek regions

(employment shares, period average, 2002-2006)

Source: Authors‟ calculations from the Amadeus database (BvD). Note: Regions have been classified into five groups using a „natural breaks‟ criterion

Figure 2 presents the same information this time split by sector

(manufacturing – non-manufacturing). As can be seen, high-concentrations of

foreign-owned activity are mainly in the regions of Attiki and Thessaloniki,

which host the two main urban centres in the country, and secondarily in the

island regions of Lesvos (for manufacturing) and the Dodecanese (for non-

manufacturing). Foreign-owned activity in the rest of the country is very sparse.


Overall, out of the 51 NUTS3 regions, in only two regions in the country does

employment in foreign-owned firms represent more than 6% of total regional

employment –while in only another 10 regions does it represent a share above

2%.

Figure 2. Foreign-owned activity by broad sector and region (employment

shares, period average, 2002-2006)

Manufacturing Non-manufacturing

Note: See note in Figure 1.

Figures 1 and 2 suggest a rather high degree of geographical concentration

of FDI around two main clusters (centred in the main cities of Athens – Attiki,

Voiotia, Evoia – and Thessaloniki – Thessaloniki, Kilkis, Pieria), with a few

„hotspots‟ of FDI activity in the remaining periphery. To examine to what extent

this visual pattern is supported by statistical evidence of spatial association

across or within regions, we performed an exploratory spatial data analysis (see

Anselin, 1995) using various definitions of neighbourliness based on distance

thresholds or on a pre-defined maximum number of neighbours.5 Figure 3

reports the results obtained from the analysis based on the 4-nearest neighbours

criterion. The maps depict the membership of regions into different types of

spatial association. The high-high cluster (black shade) includes regions that

have high shares of FDI employment and are also surrounded by regions of high

FDI shares. The high-low cluster (dark grey) includes high-FDI regions which

5 We used a range of alternative distance cut-off points as well as the k-nearest neighbour criterion

for values of k in the range {2, 8}. Using contiguity-based neighbourliness is not appropriate in the

case of Greece due to its peculiar physical geography (many island regions). In no case was the

level of global spatial association greater than 0.03 and in no case was it statistically significant

(the expected value of the Moran‟s I is around 0.16).


are surrounded by regions with low FDI shares. Similarly, the low-low and low-

high clusters include regions that have low FDI employment shares and are

surrounded by regions that also have low FDI shares (plain grey shade) or by

regions with high FDI shares (light grey stripes). Regions for which no statistical

association exists between local and neighbouring values are coloured white.

As can be seen, there is very little evidence of clustering across space. The

Moran‟s I, which measures the extent to which local outcomes correlate with

outcomes in neighbouring regions, is effectively zero; while although a few

„hotspots‟ can be identified, these are not necessarily in the places one would

expect them to appear. Specifically, for total FDI there are two main hotspots,

both located in the west (Ioannina and Etoloakarnania). These two regions are

effectively „spatial outliers‟, having relatively high shares of FDI but being

surrounded by regions with low FDI shares. Three other regions constitute

negative outliers, in the sense that they have neighbouring regions with high

values but they themselves have low shares of employment in foreign-owned

firms. Finally, two main clusters are also observable: Kozani in the north is the

centre of a low-low cluster, while Evoia in Central Greece signifies the high-

high cluster which extends southwards to Attiki and northwards to Magnisia (see

Figure 1 for the location of specific regions). Interestingly, this picture is not

replicated in either of the maps that depict the geography of spatial association

of FDI employment in the manufacturing and non-manufacturing sectors. The

Etoloakarnania outlier survives in the case of manufacturing and some new

outliers emerge in the case of non-manufacturing, but overall there is very little

consistency between the different maps, suggesting that even in cases where

local spatial association is statistically significant, the extent of clustering across

regions is rather weak.

Figure 3. Local clusters of FDI (LISA maps)

FDI:

Moran‟s I:

Total

0.0016

Manufacturing

0.0283

Non-manufacturing

-0.0200

Note: See text and footnote 7.


The high degree of geographical concentration of FDI within Greece and

its low technological content makes it plausible that, despite being a relatively

small proportion of the domestic economy, it can generate significant spillovers.

This is because spillovers often occur inside agglomerations and in a rather

localised manner (Driffield and Munday, 2001; Jordaan, 2009)6 and because the

low technological content means that problems of absorptive capacity for the

domestic firms are less likely to arise. Indeed, this is not refuted in the scant

literature that exists on the topic in Greece. The study by Dimelis and Louri

(2002) for a sample of manufacturing firms found some evidence of positive

productivity spillovers to domestic firms – but only from firms with a minority

foreign ownership. A similar effect was found by Barrios et al (2004), although

in that study the effect vanished when controls were introduced for sectoral

heterogeneity. Finally, Fotopoulos and Louri (2004) also provide indirect

evidence of positive spillovers in their analysis of foreign presence and domestic

firm growth, finding that foreign participation accelerates firm growth especially

for medium-sized firms. To our knowledge, no other study has examined the

extent and nature of FDI spillovers in Greece and no study has done so with any

attention to the geography of these spillovers. Our analysis in the remainder of

this paper seeks to fill this gap by providing unique evidence on the direction

and intensity of FDI spillovers across the Greek regions.

3. FDI spillovers across the Greek regions

As mentioned earlier, our dataset consists of firm-level data on turnover,

fixed assets and employment, organised across sectors (NACE2 and NACE4),

regions (NUTS2 and NUTS3) and years (2002-2006). Additionally, we have

aggregated the foreign ownership information at the sectoral (NACE2) and

regional (NUTS3) level to construct a variable measuring the intra-sector share

of foreign ownership in each of the 51 prefectures of Greece. As is standard in

the literature, we follow a production-function approach, where firm-level output

is made a function of each firm‟s value of fixed assets and level of employment,

adding the FDI variable as an additional regressor. Our approach implies that

investment and manning decisions are not influenced by a firms‟ own

productivity and that FDI affects a firm‟s total factor productivity but not its

level of investment or employment. Although the literature has occasionally

questioned the full validity of such assumptions (see Olley and Pakes, 1996;

Jarvocik, 2004), others have shown that the bias introduced by these

assumptions is minimal, especially in empirical studies with limited time-

horizons (Monastiriotis and Alegria, 2011).

6 See however Haskel et al (2007) for evidence against this.


Empirically, our estimating model takes the following form:

yirst = a + b1kirst + b2lirst + eirst (1)

which we later amend with the inclusion of the FDI variable and occasionally by

adding various fixed effects. Thus, our full estimating model is

yirst = a + b1kirst + b2lirst + cHrst + Rrd1 + Ssd2 + Ttd3 + Fid4 + eirst (2)

where small letters stand for logarithms, y is turnover; k is capital

(measured by fixed assets); l is employment; H is the employment share of

foreign-owned firms; R, S, T, and F are vectors containing binary dummies for

regions, sectors, time and firms, respectively; a, b1, b2, c, d1, d2, d3 and d4 are

coefficients to be estimated; i, r, s, and t index firms, regions, sectors and time,

respectively; and e is an error term. We do not restrict the coefficients b1 and b2

to add up to one, thus allowing for increasing or decreasing returns to scale. We

experiment with different definitions of the H variable (at the NUTS2 level, the

NUTS3 level, or both) and we introduce the various sets of dummy variables

selectively in alternative specifications.

Given the fact that our sample contains many dimensions (sectors,

regions, years), we start our analysis by examining the performance of our

production-function model across alternative fixed-effects specifications,

introducing gradually additional regressors to control for these dimensions.

Table 1 presents the results from this analysis. As can be seen, our base model

performs very well and the obtained factor elasticities are very robust to the

inclusion of controls for the different dimensions of our dataset.

Table 1. Production function analysis

Model:

ln(output) (1) (2) (3) (4) (5) (6) (7) (8)

ln(capital) 0.133*** 0.131*** 0.192*** 0.200*** 0.158*** 0.160*** 0.206*** 0.112***

(0.0019) (0.0019) (0.0019) (0.0019) (0.0020) (0.0020) (0.0019) (0.0026)

ln(employ-ment) 0.606*** 0.608*** 0.638*** 0.628*** 0.574*** 0.573*** 0.612*** 0.397*** (0.0039) (0.0039) (0.0036) (0.0036) (0.0039) (0.0039) (0.0036) (0.016)

Constant 5.002*** 4.901*** 4.525*** 4.440*** 4.686*** 4.715*** 4.129*** 5.564*** (0.011) (0.013) (0.038) (0.099) (0.025) (0.044) (0.10) (0.041)

Fixed effects No Time Nace2 Nace4 NUTS2 NUTS3 All Firms

& Time

Obs 98407 98407 98407 98407 98407 98407 98407 98407

R-sq 0.35 0.35 0.49 0.53 0.36 0.37 0.54 0.05

Notes: Model (8) is estimated using the Fixed Effects Within estimator. All other regressions are

estimated with OLS. NACE2 (NACE4) contains 54 (429) sectoral dummies while NUTS2

(NUTS3) contains 13 (51) regional dummies. The model of column 7 includes dummies for

NACE4, NUTS3 and time.

The coefficient on capital is rather low, but within acceptable limits, and it

increases somewhat when we add sectoral controls, which appear to control

partly for differences in capacity utilisation. The coefficient on labour is much


more stable suggesting little variation across sectors or regions in the extent of

labour hoarding. Together, the two coefficients are consistently below 1,

suggesting the presence of decreasing returns to scale in the Greek economy – a

finding consistent with the widely acknowledged inefficiency of its production

system (Bryant et al, 2001; Pagoulatos, 2003).

Importantly, adding temporal fixed effects does not affect the regression

estimates, consistent with the observation that the 2002-2006 period was a

period of relative stability for Greece. As mentioned above, sectoral controls

(either at the 2- or 4-digit of the NACE classification) raise the estimated

productivity of capital (columns 3 and 4). The influence of the regional fixed

effects (columns 5 and 6) is smaller and is again very similar for fixed effects of

different spatial scales (NUTS2 or NUTS3), suggesting that regional differences

in production technologies are minimal and smaller than differences across

sectors. Finally, introducing firm-specific fixed effects (which subsume the

regional and sectoral controls) leads to a drop in the estimated coefficients for

capital and labour, as these effects capture unobserved firm-specific

characteristics which contribute to firm output (managerial capacities,

distribution/client networks, etc).

To this basic but well performing specification we add next our FDI

variable. We experiment with different specifications of this variable for reasons

that will become clear in the discussion that follows. Table 2 presents a set of

key findings. We start by introducing a sector-specific FDI measure calculated at

the NUTS2 level (columns 1-5). When not controlling for fixed effects, of any

type, the impact of foreign firms on domestic productivity appears positive and

very significant. An increase in the employment share of foreign-owned firms by

one percentage point (e.g., from the sample average of 13% to 14%) raises

domestic productivity by 1.7%, with the effect being significant well beyond the

1% level.

Controlling for firm heterogeneity (column 2) maintains this significance

but reduces the magnitude of the estimated spillover by more than 10 times. This

clearly suggests that foreign investments concentrate in regions and sectors with

high concentrations of firms that possess productivity-enhancing unobservable

characteristics, such as good managerial practices and inter-firm networks. This

is consistent with findings elsewhere in the literature (Head et al, 1995; Hilber

and Voicu, 2010). The observed productivity spillovers increase somewhat when

we replace the firm-specific fixed effects with sector-specific ones (column 3),

but remain many times lower than those obtained through a simple OLS

estimation (column 1). Moreover, when we additionally cluster the standard

errors within sectors (column 4) the estimated spillover effect becomes


statistically not different from zero.7,8

This suggests that a large part of the self-

selection of foreign investments takes place across sectoral lines: high

productivity sectors typically attract above-average amounts of FDI.

The influence of self-selection, however, appears even stronger in its

spatial dimension. When NUTS3 fixed effects are added (col.5), the estimated

spillover effect becomes significantly negative and rather large (a rise in foreign

presence by 1 percentage point reduces domestic productivity by 0.55%). In

other words, when we control for geographical differences in productivity, the

effect of FDI turns out negative. This result, which is very consistent across

different specifications as we shall see later, has a very important implication for

the study of the spatial effects of FDI: productivity spillovers appear

misleadingly positive, largely due to the fact that foreign investments

concentrate –in the case of Greece very heavily– in regions of above-average

productivity. Net of this self-selection effect, the impact of FDI is to reduce

domestic productivity, reflecting a negative competition effect, which

presumably operates via one of the following channels: by lowering pre-existing

monopolistic rents, by creaming off skilled labour in the sector/region, or by

lowering the market share of domestic firms. All these channels are consistent

with features that are known to characterise the Greek economy, such as low

labour mobility, low effective competition within sectors and attraction of

foreign investments which are predominantly of a market-seeking type.

These results are fully replicated when using an FDI measure defined at a

much narrower geographical scale, namely the NUTS3 level (see col.6-10). The

results there are qualitatively identical to those obtained from the NUTS2-level

analysis, although in general the estimated elasticities are somewhat higher,

implying that the positive impact of foreign presence on domestic productivity is

stronger at a more localised level.

The negative spillover effect that we identify when controlling for

regional fixed effects –and thus for self-selection of foreign affiliates into high-

productivity regions– casts doubt on the conventional wisdom about the

beneficial effects of FDI but is not at odds with empirical estimates in the

international literature (Haddad and Harrison, 1993; Aitken and Harrison, 1999;

Blomström and Sjoholm, 1999; Damijan et al, 2001; Kugler, 2006;

Gorodnichenko et al, 2007). A possible factor that could account for this

negative effect, if one maintains that the overall impact of FDI should be

7 This is necessary to account for the fact that our FDI variable is measured at the sectoral level.

Clustered standard errors relax the assumption of within-cluster independence, thus allowing for

firm-specific productivities within sectors to be correlated and, by implication, for the errors to be

heteroskedastic. 8 The same result is obtained when we cluster the standard errors within regions, as well as when

we cluster within region-sector clusters (results available upon request).


T

ab

le 2

. T

he

imp

act

of

FD

I o

n d

om

esti

c p

rod

uct

ivit

y

(1

) (2

) (3

) (4

) (5

) (6

) (7

) (8

) (9

) (1

0)

(11

) (1

2)

(13

) (1

4)

Cap

ital

0

.15

6*

**

0.1

12

**

*

0.1

92

**

*

0.1

92

**

*

0.1

99

**

*

0.1

57

**

*

0.1

12

**

*

0.1

93

**

*

0.1

93

**

*

0.2

00

**

*

0.2

01

**

*

0.2

02

**

*

0.1

93

**

*

0.2

00

**

*

(0

.00

19

) (0

.00

26

) (0

.00

19

) (0

.01

6)

(0.0

15

) (0

.00

19

) (0

.00

26

) (0

.00

19

) (0

.01

6)

(0.0

15

) (0

.01

5)

(0.0

15

) (0

.01

6)

(0.0

15

)

Em

plo

ym

ent

0.5

89

**

*

0.3

96

**

*

0.6

36

**

*

0.6

36

**

*

0.6

24

**

*

0.5

88

**

*

0.3

96

**

*

0.6

35

**

*

0.6

35

**

*

0.6

24

**

*

0.6

15

**

*

0.6

06

**

*

0.6

35

**

*

0.6

24

**

*

(0

.00

38

) (0

.01

6)

(0.0

03

6)

(0.0

32

) (0

.02

8)

(0.0

03

8)

(0.0

16

) (0

.00

36

) (0

.03

2)

(0.0

28

) (0

.02

8)

(0.0

29

) (0

.03

1)

(0.0

28

)

FD

I (n

uts

2)

1.7

01

**

*

0.1

48

**

*

0.2

37

**

*

0.2

37

-0.5

49

**

*

-0

.40

7

-0.8

27

**

(0

.02

8)

(0.0

41

) (0

.03

) (0

.15

) (0

.19

)

(0.3

2)

(0.3

3)

FD

I (n

uts

3)

1

.73

0*

**

0.1

38

**

*

0.2

84

**

*

0.2

84

*

-0.4

28

**

0.6

67

**

0.3

10

(0.0

27

) (0

.04

) (0

.02

9)

(0.1

7)

(0.1

9)

(0.3

3)

(0.2

8)

FD

I(t-

1)

nu

ts3

-0.4

55

**

(0

.19

)

FD

I(t-

2)

nu

ts3

-0

.50

0*

*

(0.2

1)

Fix

ed e

ffec

ts

No

F

irm

Yea

r

NA

CE

2

Yea

r

NA

CE

2

Yea

r

NA

CE

2

NU

TS

3

Yea

r

No

F

irm

Yea

r

NA

CE

2

Yea

r

NA

CE

2

Yea

r

NA

CE

2

NU

TS

3

Yea

r

NA

CE

2

NU

TS

3

Yea

r

NA

CE

2

NU

TS

3

Yea

r

NA

CE

2

Yea

r

NA

CE

2

NU

TS

3

Yea

r

Est

imat

ion

met

ho

d

OL

S

Wit

hin

O

LS

O

LS

+

cl(n

ace2

)

OL

S +

cl(n

ace2

) O

LS

W

ith

in

OL

S

OL

S +

cl(n

ace2

)

OL

S +

cl(n

ace2

)

OL

S +

cl(n

ace2

)

OL

S +

cl(n

ace2

)

OL

S +

cl(n

ace2

)

OL

S +

cl(n

ace2

)

Co

nst

ant

4.7

03

**

*

5.5

46

**

*

4.4

30

**

*

4.4

30

**

*

4.2

40

**

*

4.7

02

**

*

5.5

48

**

*

4.4

28

**

*

4.4

28

**

*

4.2

45

**

*

4.3

06

**

*

4.4

44

**

*

4.4

30

**

*

4.2

39

**

*

-0

.01

2

-0.0

42

-0.0

39

-0.0

75

-0.1

2

-0.0

12

-0.0

42

-0.0

39

-0.0

74

-0.1

2

-0.1

1

-0.1

1

-0.0

75

-0.1

2

Ob

serv

atio

ns

98

40

7

98

40

7

98

40

7

98

40

7

98

40

7

98

40

7

98

40

7

98

40

7

98

40

7

98

40

7

79

80

1

60

04

5

98

40

7

98

40

7

R-s

qu

ared

0

.37

0.0

5

0.4

9

0.4

9

0.5

0

.37

0.0

5

0.4

9

0.4

9

0.5

0

0.4

9

0.4

9

0.4

9

0.5

0

Note

s: E

stim

ated

sta

nd

ard

err

ors

in

par

enth

eses

. **

*.

**

and

* i

ndic

ate

sign

ific

ance

at

the

1,

5 a

nd

10

% l

evel

s. E

stim

atio

ns

wher

e st

and

ard e

rro

rs a

re

clust

ered

wit

hin

NA

CE

2 s

ecto

rs a

re i

ndic

ated

wit

h „

cl(n

ace2

)‟;

„Wit

hin

‟ is

the

fixed

eff

ects

pan

el d

ata

esti

mat

or.


positive, is the possibility that the beneficial effects of FDI take time to

materialise, i.e. occur with some hysterisis (Monastiriotis and Alegria, 2011).

If this hypothesis is valid, then we should observe a contemporaneous negative

association between foreign presence and domestic productivity but a more

positive relation between current domestic productivity and past values of FDI.

Although year-to-year variation in the share of foreign presence in the Greek

economy is limited, which somewhat problematises the identification of this

mechanism in our data, our estimations that test the hysterisis hypothesis

(columns 11 and 12) do not seem to support this assumption: the estimated

spillovers remain negative when we replace the contemporaneous FDI variable

with its one- and two-year lags. In fact, the magnitude of the negative FDI effect

is actually increasing, suggesting if anything a deterioration of domestic

productivity as a response to foreign presence over time. It thus appears that

domestic firms do not adapt (at least not in a two-year horizon) to the negative

shock of foreign presence and continue to suffer from the increased competition

generated by the foreign affiliates. Again, this is consistent with the view of

Greece as an economy that lacks dynamism and where competition is largely a

zero-sum game which does not lead to market expansion.

Returning to the point that we raised earlier about the geographical scale

at which the positive and negative spillovers of FDI occur, in the last two

columns of Table 2 we include simultaneously two FDI variables, measured at

two different geographical scales (NUTS2 and NUTS3), alternatively excluding

and then including our controls for regional fixed effects. In both cases, an

interesting pattern emerges: FDI spillovers appear negative at the wider

geographical scale but are consistently positive at the more localised level.9 This

suggests that locally concentrated FDI helps the performance of domestic firms,

especially in comparison to the performance of similar firms in other NUTS3

regions (since the estimated NUTS3 spillover is stronger and larger when not

including regional fixed effects). At the same time, concentration of FDI in

neighbouring areas, within a local economy‟s administrative region, has an

absolute negative effect on the performance of domestic firms. This offers an

important insight into the workings of FDI spillovers in Greece. Positive FDI

spillovers, presumably due to both pecuniary (demand) and technology effects

(demonstration, imitation), do exist, but they are very localised. Indeed, these

benefits do not diffuse to wider geographical scales and thus at the regional (and

national) level the competition and market capture effect of FDI dominates.

Therefore, despite the localised benefits, the overall effect of FDI on domestic

9 The NUTS2 and NUTS3 spillover coefficients reported in columns 13 and 14 are not

simultaneously significant statistically. They are however jointly significant in each of the

regressions and also significantly different from each other. When replicating these regressions

without clustering the standard errors all coefficients are highly significant even at the 0.1% level.


productivity, when taking into account the tendency of foreign investments to

self-select into high-productivity regions and sectors, is negative.

4. Spatial and functional heterogeneity of FDI spillovers

The literature on FDI productivity spillovers has often found that

spillovers do not accrue homogeneously across different types of firms and

sectors but are rather dependent on specific firm characteristics such as size,

technology content and sector of economic activity. A similar argument can be

made about the heterogeneity of FDI effects across space, especially under the

light of our preceding discussion and findings. In this section we explore these

two issues, starting with the functional dimension. Table 3 reports the results

from a set of regressions where we split the sample across sectoral

characteristics (manufacturing versus non-manufacturing, high-tech sectors

versus low-tech sectors), firm sizes (large, medium, small), and firm-specific

technological content (high/low technology gap10

). We perform this analysis for

two alternative specifications, first excluding (top panel) and then including

regional fixed effects (bottom panel).

The results are particularly revealing. Although in virtually all cases the

pattern of positive localised spillovers and negative overall spillovers is

maintained (with the exception of the results obtained for large firms, as

discussed below), there are important variations in the effects observed for

different firm types. The impact of FDI is much more heightened outside the

manufacturing sector: the positive localised (NUTS3) effect is stronger, while

the negative diffused (NUTS2) effect is also more pronounced. In contrast, the

impact of FDI in the manufacturing sector is statistically weak (although the

introduction of only one FDI variable at a time makes the results consistent with

our earlier analysis). This is true both for including or excluding regional fixed

effects. In the latter case, the negative diffused effect becomes larger, while the

positive localised effect is smaller. This suggests that in non-manufacturing

sectors foreign firms self-select into high-productivity localities but,

interestingly, away from localities that are neighbouring to high productivity

ones.11

In a somewhat similar fashion, localised FDI spillovers appear stronger

(more positive) in high-tech sectors, especially when we do not control for self-

selection, and the diffused spillover effect appears more negative, especially

10 This is measured as the distance of each domestic firm from the productivity frontier of its

sector, proxied by the level of productivity achieved by the most productive foreign-owned firm in

the sector nationally. 11 For the manufacturing sector the opposite pattern is observed, with foreign firms self-selecting

into broader regions of high productivity but not necessarily into the localities with the highest

productivity within these broad regions. The results here, however, lack statistical significance and

thus this interpretation is tentative.


when we do control for self-selection. Although the results for the low-tech

sectors are of a similar nature, they are more modest and at the margin of

statistical significance. Our findings, however, differ markedly when we split

our sample by firm size. For large firms, the presence of foreign-owned affiliates

within the same locality appears to produce negative, not positive, productivity

effects, especially when we control for self-selection. In contrast, it is the

diffused spillovers that turn out positive. This suggests an interesting property

for large firms in Greece: co-location with foreign affiliates seems to hamper

their performance, presumably as large firms have less to gain from

demonstration effects and pecuniary spillovers accruing from their foreign-

owned competitors; but the agglomeration of foreign firms in the wider region

outside the domestic large firms‟ own locality has a positive effect on their

performance. The absence of localised benefits in the presence of wider-scale

ones seems to suggest that foreign-firm concentration tends to generate a

market-creation effect for large firms, which is not operational in the case of

medium and small firms within the same sector. For the latter, and especially for

medium-sized firms, the main (and only) benefit is from the presence of foreign

affiliates within the local economy, while the wider-scale effect is consistently

and very strongly negative. An obvious interpretation of this finding is that

smaller firms do not have the reach to capture the benefits from the market-

creation effect at wider geographical scales. Medium-sized firms appear able to

internalise successfully some of the positive spillovers of foreign participation in

the local economy, while very small firms lack the absorptive capacity to do so,

which would explain why the estimated localised spillover effect fails to reach

acceptable levels of statistical significance for these firms.

The level of productivity (technology gap) of domestic firms does not

seem to be a crucial factor for the realisation of FDI spillovers. Firms with a

lower technological gap appear to benefit more from the presence of foreign

firms in the local economy and to suffer less from the agglomeration of foreign

affiliates at the wider geographical scale outside the local economy – although

self-selection appears to be more important here than in the case of firms with a

lower technology content (higher technology gap).12

When controlling for

selection, the latter appear to be better placed to reap the benefits of co-location

(technology transfers and pecuniary effects), although they remain more

susceptible to suffering from competition with foreign-owned firms at a wider

geographical scale (market capture effect).

What do these patterns imply for the geography of productivity spillovers

accruing from the geographical concentration of foreign-owned firms across the

Greek economy?

12 This implies that foreign firms tend to locate in areas with higher concentrations of high

technology firms within any given sector – a finding which is consistent with widespread evidence

in the literature concerning the location of FDI.


Tab

le 3

. Fu

nct

ion

al h

eter

ogen

eity

of

FD

I sp

illo

vers

M

anu-

fact

urin

g N

on-m

anuf

. H

igh-

tech

Lo

w-t

ech

Larg

e SM

Es

Med

ium

Sm

all

Low

tech

n. g

ap

Hig

h te

chn.

gap

W

ITH

OU

T RE

GIO

NAL

DU

MM

IES

Capi

tal

0.22

4***

0.

186*

**

0.19

2***

0.

193*

**

0.25

2***

0.

175*

**

0.18

6***

0.

148*

**

0.13

3***

0.

134*

**

(0

.021

) (0

.019

) (0

.025

) (0

.019

) (0

.025

) (0

.015

) (0

.017

) (0

.019

) (0

.016

) (0

.019

) Em

ploy

men

t 0.

633*

**

0.63

2***

0.

644*

**

0.63

3***

0.

690*

**

0.53

2***

0.

729*

**

0.30

7***

0.

716*

**

0.79

2***

(0.0

27)

(0.0

40)

(0.0

52)

(0.0

38)

(0.0

38)

(0.0

28)

(0.0

42)

(0.0

26)

(0.0

29)

(0.0

41)

FDI (

nuts

2)

-0.0

323

-0.9

15*

-0.5

14

-0.3

50

0.56

2***

-0

.705

**

-0.6

18*

-0.8

13*

0.05

27

-0.5

05

(0

.27)

(0

.46)

(0

.60)

(0

.37)

(0

.18)

(0

.32)

(0

.33)

(0

.43)

(0

.16)

(0

.36)

FD

I (nu

ts3)

0.

398

1.13

**

1.35

**

0.49

1 -0

.111

0.

853*

* 0.

810*

* 0.

810*

0.

549*

**

0.50

7

(0.2

8)

(0.5

0)

(0.5

8)

(0.3

8)

(0.1

5)

(0.3

4)

(0.3

3)

(0.4

7)

(0.1

6)

(0.3

5)

Cons

tant

5.

106*

**

4.48

7***

3.

075*

**

4.45

2***

3.

934*

**

4.64

3***

4.

028*

**

5.08

4***

4.

797*

**

1.69

6***

(0.0

49)

(0.0

88)

(0.1

9)

(0.0

82)

(0.1

4)

(0.0

70)

(0.1

1)

(0.0

73)

(0.0

80)

(0.0

61)

Obs

erva

tions

26

224

7218

3 18

563

7984

4 20

520

7788

7 53

742

2414

5 48

366

5004

1 R-

squa

red

0.55

0.

47

0.47

0.

49

0.52

0.

34

0.34

0.

22

0.73

0.

58

IN

CLU

DIN

G R

EGIO

NAL

DU

MM

IES

Capi

tal

0.23

6***

0.

191*

**

0.19

5***

0.

199*

**

0.26

0***

0.

181*

**

0.19

5***

0.

151*

**

0.14

1***

0.

135*

**

(0

.019

) (0

.017

) (0

.023

) (0

.018

) (0

.023

) (0

.014

) (0

.015

) (0

.018

) (0

.017

) (0

.018

) Em

ploy

men

t 0.

620*

**

0.62

0***

0.

634*

**

0.62

1***

0.

676*

**

0.52

3***

0.

712*

**

0.30

7***

0.

701*

**

0.79

3***

(0.0

25)

(0.0

35)

(0.0

48)

(0.0

34)

(0.0

35)

(0.0

25)

(0.0

41)

(0.0

27)

(0.0

26)

(0.0

40)

FDI (

nuts

2)

-0.1

91

-1.5

2***

-0

.819

***

-0.7

65*

0.29

5 -1

.16*

**

-1.1

6***

-1

.04*

**

-0.2

07

-0.6

42*

(0

.27)

(0

.33)

(0

.25)

(0

.39)

(0

.21)

(0

.31)

(0

.33)

(0

.38)

(0

.16)

(0

.38)

FD

I (nu

ts3)

0.

0598

0.

661*

* 0.

536*

* 0.

254

-0.4

80**

0.

499*

0.

506*

0.

446

0.27

5*

0.59

8*

(0

.29)

(0

.30)

(0

.23)

(0

.33)

(0

.19)

(0

.26)

(0

.29)

(0

.36)

(0

.15)

(0

.34)

Co

nsta

nt

4.58

9***

4.

461*

**

2.36

6***

4.

265*

**

4.05

2***

4.

419*

**

3.67

3***

5.

064*

**

4.67

1***

1.

566*

**

(0

.093

) (0

.12)

(0

.34)

(0

.13)

(0

.18)

(0

.11)

(0

.12)

(0

.16)

(0

.10)

(0

.12)

Obs

erva

tions

26

224

7218

3 18

563

7984

4 20

520

7788

7 53

742

2414

5 48

366

5004

1

R-sq

uare

d 0.

56

0.48

0.

49

0.49

0.

52

0.35

0.

34

0.23

0.

73

0.58

Not

es:

Est

imat

ed s

tand

ard

erro

rs i

n pa

rent

hese

s. *

**.

** a

nd *

ind

icat

e si

gnif

ican

ce l

evel

s at

the

1,

5 an

d 10

%.

All

reg

ress

ions

inc

lude

tim

e an

d se

ctor

dum

mie

s, w

ith

stan

dard

err

ors

clus

tere

d at

the

NA

CE

2 le

vel.


We examine this by replicating the regressions presented in columns 5

and 10 of Table 2, this time interacting the FDI variable with a set of regional

(alternatively, NUTS2 and NUTS3) dummies.13

This provides us with a full set

of region-specific estimates of the productivity effects of FDI on domestic firms.

It is of course difficult to report the full set of obtained coefficients in tabular

form (in the NUTS3 analysis, this set contains 51 region-specific spillover

coefficients). Instead, in Figure 4 we offer a visualisation of the magnitude and

geography of these effects.

Figure 4. Region-specific estimates of local FDI spillovers

Without local/regional fixed-effects Including local/regional fixed effects

Notes: Estimated coefficients derived from an extension of models 5 and 10 of Table 2, which

includes interaction terms between the FDI variable and the regional dummies. See the text for

more details.

13 We can not replicate this analysis for the models presented in columns 13 and 14 of Table 2, as

the NUTS2-level effects are absorbed completely by the NUTS3-level effects when both are

included in the same model.

Reg

ion

-sp

ecif

ic s

pil

lov

ers

(NU

TS

2)

Lo

cal-

spec

ific

sp

illo

ver

s (N

UT

S3

)

Lo

cal-

spec

ific

sp

illo

ver

s (N

UT

S3

)


The top panel depicts the results obtained from the NUTS3-level analysis

(corresponding to the regression of column 10 in Table 2), while the bottom

panel presents the results of the NUTS2-level analysis. The right-hand side maps

correspond to regressions that include locational fixed effects (regional or local),

which take into account the self-selection of foreign firms into high-productivity

regions.

As can be seen, there is significant variation both in the size and, more

importantly, in the direction of the estimated effects. At the NUTS2 level, when

not controlling for self-selection (bottom-left map), this variation is less

heightened and the overall picture appears to be one of a core-periphery pattern.

Peripheral regions in the north-east, west and south appear unable to internalise

positive productivity spillovers, thus ending up with a net loss in their

performance. In contrast, more central regions, especially in central and western

mainland Greece, experience net gains from the presence of foreign-owned

firms.

Controlling for self-selection, however, completely overturns this picture:

in this case, the direct productivity effect of FDI is negative in all regions

(consistent with the findings reported in column 5 of Table 2), but with one

important exception. The region of Eastern Macedonia and Thrace (EMT:

located in north-east mainland Greece) appears now to benefit from positive

spillovers, suggesting that self-selection in this region operates in the opposite

direction. The structure of incentives operating in this region through the

country‟s Incentives Laws (Filippaios and Kottaridi, 2004) may have a big role

to play here, as foreign firms may locate there not because the region offers a

high concentration of more productive firms but rather because the structure of

incentives provided by the government attracts high-productivity foreign

affiliates to this low-productivity region. In any case, the issue of regional

incentives aside, our results suggest that had average productivity in this region

been the same as the average productivity nationally, the effect of foreign

presence in the region would have been to raise the overall productivity of the

domestic firms located there.

The NUTS3-level analysis (top panel) suggests that this is essentially due

to two more localised effects: a negative effect in Xanthi (located in the middle

of the EMT region), which after controlling for selection turns mildly positive,

and a mildly positive effect in Evros (the eastern-most prefecture of the EMT

region), which after controlling for selection becomes even stronger. Besides

this, self-selection seems to operate more strongly (and to result to a negative

overall effect of FDI) in the prefectures of Attiki, Thessaloniki and Larissa –the

regions hosting three of the five largest cities in Greece– and less strongly in the

case of Etoloakarnania (the western-most region of central Greece). FDI

spillovers are invariably negative (irrespective of controls for region-specific

fixed effects) in Pella (in the north), Kerkyra (the western-most island) and


Chania and Irakleio (both in the southern-most island of Crete); while inverse

self-selection (into low productivity regions) appears, besides Xanthi and Evros,

in the cases of Argolida (south-west of Athens), Lasithi (in eastern Crete),

Trikala (located north of central Greece), Ipeiros (west of Trikala) and the

Dodecanese (the south-eastern island complex). Interestingly, the estimated

spillovers of FDI are persistently negligible in the prefectures of Lesvos (north-

east islands), Rethymno (in Crete), Voiotia (north of Athens), Magnisia (in

central-east mainland Greece) and Achaia and Korinthos (both in northern

Peloponese)14

, while the most positive FDI effect at the NUTS3 level, which

appears also independent of self-selection, is observed in the case of Preveza –

whose predominantly agricultural economy (representing some 30% of total

employment) has attracted in the past some modest, but highly concentrated,

foreign investment in the manufacture of wood and wood products.

5. Discussion

This paper has sought to contribute to filling an important gap in the

literature of FDI-induced spillovers, by providing a detailed and novel analysis

of the spatial heterogeneity of the productivity effects of foreign-owned firms at

the local and regional level. Our results confirm that FDI tends to concentrate in

a limited number of locations, self-selecting into regions and sectors of high

productivity. It thus acts to heighten existing spatial imbalances, as the

productive capacity of the most developed regions is strengthened and the

relative performance of regions located in the economic periphery deteriorates.

Nevertheless, although this effect on the spatial structure is important, our

analysis has found that FDI does not raise the productivity of domestic firms,

neither contemporaneously nor in a longer time-horizon. In this sense, the

concentration of FDI in the most developed regions in the country is not a

hindrance to regional growth and convergence for the less well-off regions. This

is consistent with the scant evidence in the literature about the localised effects

of FDI.

Besides this, our analysis has shown further that the productivity

spillovers of FDI exhibit substantial heterogeneity across space, even after

controlling for regional differences in the volume and sectoral composition of

FDI. To our knowledge, this is a unique finding in the literature. Moreover, it

has very important policy implications, pointing to a strong need for FDI-

attracting policies to incorporate a clear regional dimension. This is because if,

as it seems, FDI is not equally beneficial (or harmful) across the national

economic space, maximising the benefits of FDI at the aggregate/national level

necessitates paying specific attention to the set of endogenous (e.g., average firm

sizes) and exogenous (e.g., proximity to main agglomerations) locational

14 This is despite the fact that the latter four are rather highly industrialised regions.


characteristics that influence local abilities to benefit from FDI spillovers. In

other words, it requires policies that are spatially targeted and selective.

Our analysis has identified a number of factors that condition the

externalities generated by the presence of foreign-owned activity. Some of these,

concerning firm- and sector-specific characteristics, have already been identified

in the literature and the evidence presented here has lent further support to them.

Two other factors identified in our analysis, however, are novel in character and

perhaps also more important for understanding the spatial processes that underlie

FDI spillovers. The first has to do with the extent of urbanisation and the size of

the recipient economy. Our results indicate that FDI spillovers are negative

mainly in regions that host the largest urban areas (Athens, Thessaloniki,

Irakleio, Larissa). In contrast, they are positive, even after controlling for

selection, in smaller and more peripheral regions. Whether this signifies an

adverse agglomeration effect or something qualitatively different15

, its policy

implications are clear. At least in the case of Greece, the spatial selectivity of

policies seeking to maximise the productivity effects of FDI should be such so as

to direct foreign investments towards less dynamic, less urbanised and less

competitive regions in the country. This is particularly relevant for Greece today,

as the country is ready to embark on a new phase of FDI promotion to deal with

the acute investment problems that it faces following the fiscal crisis and the

austerity measures that were implemented to address it.

The second factor concerns geographical proximity and the scale of the

spillovers. Domestic firms tend to enjoy positive FDI spillovers at the very

localised level (with the exception of large domestic firms, which operate at a

different scale), even after controlling for self-selection of FDI firms into high-

productivity areas. The overall effect remains negative, but this is due to a very

strong negative effect on local productivity coming from the location of foreign

investments in neighbouring regions. The implication of this finding is of

paramount importance and has foregone the attention in most of the FDI

literature so far: not only is the effect of FDI spatially heterogeneous or

conditioned on specific firm, sectoral, and regional characteristics, it is moreover

dependent on geographical proximity.

To us, this seems to suggest that different mechanisms are in operation at

different geographical scales – at least in the case of Greece. Although we

cannot provide conclusive evidence to support this interpretation, it appears that

processes of technology diffusion and learning are very localised, taking place at

the prefectural level within NUTS3 areas. In contrast, at wider geographical

scales the effect that dominates is a negative competition effect of market

capture and demand siphoning, where foreign-owned firms limit the market size

15 For example, it is consistent with the observation that FDI spillovers tend to be weaker in areas

exposed to high domestic and international competition, because firms in such areas have already

acquired the technological features that foreign-owned firms are believed to incorporate.


of the domestic firms and thus push upwards their average production costs and

reduce their productivity – as domestic firms find it difficult to adjust either

positively (for example, through product differentiation and expansion to new

markets) or negatively (through disinvestment and downsizing) to the foreign

presence. This may be a feature unique to Greece, as the country is known to

have rather inflexible industrial relations and inefficient managerial practices,

but our sense is that it may be true, perhaps to different extents, also in other

countries, at least in cases where significant spatial differences exist in the

competitiveness and extroversion of local firms.

The extent to which this is true, and the particular mechanics under which

this process takes effect (e.g., the role of agglomeration, openness, industrial

diversity, etc), is something that we could not address in this paper and that

future research needs to address. For what concerns the present analysis, the

main conclusions that we can draw are the following. Foreign investments have

inequitable location patterns that can intensify existing spatial and sectoral

asymmetries. In economies such as that of Greece, however, such investments

do not generate positive productivity spillovers, especially in more developed

regions. Therefore, their overall impact on relative regional performance and

cross-regional convergence is not detrimental. Positive spillovers, when they

exist, are very localised and dominated by wider-area negative market-capture

effects. It follows that a successful FDI promotion and regional development

policy is not a policy that maximises the FDI flows accruing to the country but

one that addresses effectively two key issues: the location of FDI within the

national economic space and the conversion of negative competition into

economic extroversion and market expansion.

References

Aitken, B.J. and Harrison, A. (1999), Do domestic firms benefit from direct foreign

investment? Evidence from Venezuela, American Economic Review, Vol.89, No.3, pp.

605-618.

Anselin, L. (1995), Local Indicators of Spatial Association – LISA, Geographical

Analysis, Vol. 27, pp. 93-115.

Apergis, N. and Katrakylidis, C. (1998), Does Inflation Uncertainty Matter in Foreign

Direct Investment Decisions? An Empirical Investigation for Greece, Spain, and

Portugal, Rivista Internazionale di Scienze Economiche e Commerciali, Vol. 45, No. 4,

pp. 729-744.

Barbosa, N. and Louri, H. (2002), On the Determinants of Multinationals' Ownership

Preferences: Evidence from Greece and Portugal, International Journal of Industrial

Organization, Vol. 20, No. 4, pp. 493-515.


Barrios, S., Dimelis, S., Louri, H. and Strobl, E. (2004), Efficiency Spillovers from

Foreign Direct Investment in the EU Periphery: A Comparative study of Greece, Ireland

and Spain, Review of World Economics, Vol. 140, No. 4, pp. 688-705.

Bitzenis, A., Tsitouras, A. and Vlachos, V. (2007), Motives for FDI in a Small EMU

Member State: the case of Greece, Journal of Economics and Business, Vol. 10, No. 2,

pp. 11-42.

Blalock, G. and Gertler, P.J. (2008), Welfare gains from foreign direct investment

through technology transfer to local suppliers, Journal of International Economics, Vol.

74(2), pp. 402-421.

Blomstorm, M. and Kokko, A. (1998), Multinational corporations and spillovers,

Journal of Economic Surveys, Vol. 12, pp. 247-277.

Blomström, M. and Sjoholm, F. (1999), Technology Transfer and Spillovers: Does Local

Participation Matter?, European Economic Review, Vol. 43, pp. 915-923.

Bryant, R., Garganas, N. and Tavlas, G. (eds.) (2001), Greece’s Economic Performance

and Prospects, Bank of Greece/The Brookings Institution, Athens.

Cantwell, J. and Iammarino, S. (2001), EU Regions and Multinational Corporations:

Change, stability and Strengthening of Technological Advantages, Industrial and

Corporate Change, Vol. 10, No. 4, pp. 1007-1037.

Caves, R.E. (2007), Multinational enterprise and economic analysis, Cambridge,

Cambridge University Press.

Damijan, J., Majcen, B., Knell, M. and Rojec, M. (2001), The Role of FDI, Absorptive

Capacity and Trade in Transferring Technology to Transition Countries, UN Economic

Commission for Europe, Geneva.

De Propris, L. and Driffield, N. (2006), The importance of clusters for spillovers from

foreign direct investment and technology sourcing, Cambridge Journal of Economics,

Vol. 30(2), pp. 277-291.

Dimelis, S. and Louri, H. (2002), Foreign ownership and production efficiency: a

quantile regression analysis, Oxford Economic Papers, Vol. 54, pp. 449-469.

Driffield, N. and Munday, M. (2001), Foreign Manufacturing, Regional Agglomeration

and Technical Efficiency in UK Industries: A Stochastic Production Frontier Approach,

Regional Studies, Vol. 35, No. 5, pp. 391–399.

Dunning, J.H. and Lundan, S.M. (2008), Multinational enterprises and the global

economy, Cheltenham, UK: Edward Elgar.

Filippaios, F. and Kottaridi, C. (2004), Investment patterns and the competitiveness of

Greek regions, Review of Urban and Regional Development Studies, Vol. 16, No. 2, pp.

93-112.

Fotopoulos, G. and Louri, H., (2004), Firm Growth and FDI: Are Multinationals

Stimulating Local Industrial Development?, Journal of Industry, Competition and Trade,

Vol. 4, pp. 163-189.


Georganta, Z., Manos, K., Notis, B., Robolis, S., and Sakkas, D. (1986), Foreign

Investment in Greece, Athens: Centre for Planning and Economic Research, (in Greek).

Gersl, A., Rubene, I. and Zummer, T. (2007), Foreign direct investment and productivity

spillovers: Updated evidence from Central and Eastern Europe, Czech National Bank

Working Papers, No. 8.

Georgopoulos, A. and Preusse, H. G. (2006), European Integration and the Dynamic

Process of Investments and Divestments of Foreign TNCs in Greece, European Business

Review, Vol.18, No. 1, pp. 50-59.

Girma, S. (2005), Technology transfer from acquisition FDI and the absorptive capacity

of domestic firms: An empirical investigation, Open Economies Review, Vol. 16, No.2,

pp. 175-187.

Gorodnichenko, Y., Svejnar, J. and Terrell, K. (2007), When Does FDI Have Positive

Spillovers? Evidence from 17 Emerging Market Economies, IZA Discussion Paper

Series, No. 3079.

Guimaraes, P., Figueiredo, O., and Woodward, D., (2000), Agglomeration and the

Location of Foreign Direct Investment in Portugal, Journal of Urban Economics, Vol.

47, pp. 115-135.

Haddad, M. and Harrison, A. (1993), Are There Positive Spillovers from Direct Foreign

Investment? Evidence from Panel Data for Morocco, Journal of Development

Economics, Vol. 42, pp. 51-74.

Halpern, L. and Muraközy, B. (2007), Does distance matter in spillover?, Economics of

Transition, Vol. 15, pp. 781-805.

Haskel, J., Pereira, S. and Slaughter, M. (2007), Does Inward Foreign Direct Investment

Boost the Productivity of Domestic Firms?, Review of Economics and Statistics, Vol. 89,

No. 3, pp. 482-496.

Head, K., Ries, J. and Swenson, D. (1995), Agglomeration benefits and location choice:

evidence from Japanese manufacturing investments in the United States, Journal of

International Economics, Vol. 38, pp. 223-247.

Hilber, C. and Voicu, I. (2010), Agglomeration economies and the location of foreign

direct investment: Empirical evidence from Romania, Regional Studies, Vol. 44 (3), pp.

355-371.

Jarvocik, B. (2004), Does Foreign Direct Investment Increase the Productivity of

Domestic Firms? In Search of Spillovers through Backward Linkages, American

Economic Review, Vol. 94, pp. 605-627.

Jordaan, J.A. (2005), Determinants of FDI-induced externalities: New empirical

evidence from Mexican manufacturing industries, World Development, Vol. 33, No. 12,

pp. 2103-2118.

Jordaan, J.A. (2008), Regional foreign participation and externalities: new empirical

evidence from Mexican regions, Environment and Planning, Vol. 40, No.12, pp. 2948-

2969.


Jordaan, J.A. (2009), Foreign direct investment, agglomeration and externalities,

Surrey: Ashgate Publishing Limited.

Kinoshita, Y. (2000), R&D and technology spillovers through FDI: Innovation and

absorptive capacity, William Davidson Institute Working Papers, No. 349.

Kokkinou, A. and Psycharis, I. (2004), Foreign Direct Investments, Regional Incentives

and Regional Attractiveness in Greece, paper presented at the 44th ERSA Congress

(Porto, Portugal).

Kokko, A. (1994), Technology, market characteristics and spillovers, Journal of

Development Economics, Vol. 43, No. 2 , pp. 279-293.

Kugler, M. (2006), Spillovers from Foreign Direct Investment: Within or Between

Industries?, Journal of Development Economics, Vol. 80, pp. 444-477.

Louri, H., Papanastassiou, M. and Lantouris, J. (2000), FDI in the EU Periphery: A

Multinomial Logit Analysis of Greek Firm Strategies, Regional Studies, Vol. 34, No. 5,

pp. 419-427.

Merlevede, B. and Schoors, K. (2005), How to catch foreign fish? FDI and privatisation

in EU accession countries, Working Papers of Faculty of Economics and Business

Administration, Ghent University, Belgium.

Monastiriotis, V. and Alegria, R. (2011), Origin of FDI and domestic spillovers:

evidence from Greek and European FDI in Bulgaria, Review of Development Economics,

Vol. 15, No. 2, (forthcoming).

Moran, T., Graham, E. and Blomström, M. (eds.) (2005), Does Foreign Direct

Investment Promote Development?, Institute for International Economics and Center for

Global Development, Washington, D.C.

Mullen, J. and Williams, M. (2007), Foreign Direct Investments and Regional

Productivity Spillovers in US Manufacturing, Review of Urban and Regional

Development Studies, Vol. 19, No. 3, pp. 185-196.

Olley, S. and Pakes, A. (1996), The Dynamics of Productivity in the

Telecommunications Equipment Industry, Econometrica, Vol. 64, pp. 1263-1297.

Pagoulatos, G. (2003), Greece’s New Political Economy: State, Finance and Growth

from Postwar to EMU, Oxford University Press, Oxford.

Pantelidis, P. and Nikopoulos, E. (2008), FDI attractiveness in Greece, International

Advances Economic Research, Vol. 14, pp. 90-100.

Papazoglou, C.E., (2001), Regional economic integration and FDI inflows: the European

experience, Bank of Greece, Economic Bulletin, No. 17, (in Greek).

Resmini, L., (2008), The location of multinational enterprises in Central and Eastern

European countries, in Krieger-Boden, Morgenroth and Petrakos, The Impact of

European Integration on Regional Structural Change and Cohesion, Routledge.

Smeets, R. (2008), Collecting the Pieces of the FDI Knowledge Spillovers Puzzle, World

Bank Research Observer, Vol. 23, No. 2, pp. 107-138.


UNCTAD (2004), World Investment Report, United Nations Conference on Trade and

Development, Geneva.

UNCTAD (2009), World Development Report, Annex B, United Nations Conference on

Trade and Development, Geneva.

Documents

Does FDI promote regional development? Evidence …in 2008 manufacturing accounted for some 33% of the total stock of inward FDI, 3 Ambiguity also exists with regard to the location